Background: The Risedronate and Alendronate (REAL) cohort study provides unique comparative effectiveness data for real world bisphosphonate treatment of osteoporosis.
Objective: The objective of this analysis was to assess the cost-effectiveness of risedronate compared to generic alendronate in Germany applying the REAL effectiveness data.
Materials and Methods: A validated Markov model of osteoporosis was populated with REAL effectiveness data and German epidemiological, cost, and utility data. To estimate the impact of therapy on hip fractures, costs, and quality adjusted life years (QALYs), the analysis included women ≥65 years, treated with risedronate or alendronate and followed for 4 additional years. Country-specific data included population mortality, fracture costs, and annual drug costs, using a German social insurance perspective. Costs and outcomes were discounted at 3%. A differential hip fracture relative risk reduction of 43% was applied to risedronate vs. alendronate.
Results: The model predicted that treatment with risedronate would result in fewer hip fractures and more QALYs at a reduced cost (savings of €278 per treated woman) compared to treatment with generic alendronate. Sensitivity analysis assuming 2 years of treatment and equivalence of effect after 1 year show cost savings as well (€106 per treated woman).
Discussion: Whereas previous economic evaluations involving bisphosphonates have mainly relied on efficacy data from noncomparative clinical trials, this study's strength is in the use of comparative effectiveness data from one data source. The magnitude of the cost savings observed were sensitive to alternative assumptions regarding treatment duration, therapy discontinuation and cost of generic alendronate.
Conclusions: Based on “real world” data the analysis supports the first line use of risedronate for the treatment of osteoporotic women in Germany.
If you can't find a tool you're looking for, please click the link at the top of the page to go "Back to old version". We'll be adding more features regularly and your feedback is important to us, so please let us know if you have comments or ideas for improvement.
In Germany, osteoporosis affects approximately 7.8 million people or 26% of the population 50 years of age or older . The total direct costs attributable to osteoporosis in the year 2003 were estimated at €5.4 billion, accounting for 3.5% of all health-care expenditures by the social health insurance fund . Inpatient costs relating to fracture treatment accounted for more than half (56%) of the total direct costs . The prevalence and economic burden of osteoporosis in Germany are expected to increase significantly over the next decade, due to the aging of the population. Despite these estimates, data suggest a low awareness of osteoporosis among physicians and the general population .
Bisphosphonates have been well established through randomized controlled trials (RCTs) as an efficacious treatment for postmenopausal women at high risk of fracture [3–9]. In Germany, risedronate and alendronate are the most commonly used bisphosphonates accounting for more than 90% of the bisphosphonate market share. In a recent systematic review of the cost-effectiveness of bisphosphonates, Fleurence et al.  identified 16 studies conducted in women with low bone mineral density (BMD) [11–25]. Two studies included a comparison of risedronate and alendronate [15,16]; both found risedronate to be cost-effective relative to alendronate in patients aged 60 to 69 years of age with or without previous vertebral fracture. Nevertheless, these studies were limited in that they combined efficacy data from independent sources in absence of head-to-head clinical efficacy data.
Although there remains no direct comparison of risedronate and alendronate in RCTs, new comparative data now exist from a medical claims database study. The RisedronatE and ALendronate (REAL) study found that patients treated with risedronate had an 18% (95% confidence interval [CI]: 0.02–0.32) lower incidence of nonvertebral fractures and a 43% (95% CI: 0.13–0.63) lower incidence of hip fractures at 12 months, compared to alendronate . Beyond conventional RCTs, data from medical claims databases can provide additional evidence of treatment effectiveness in a broad population of patients who are prescribed the drug according to actual clinical practice and provide a unique opportunity to conduct comparative cost-effectiveness analyses based on effectiveness data. Recent policy initiatives reinforce that decision-makers are increasingly recognizing the importance of incorporating real-world evidence into health care and reimbursement decision-making .
Recently, Grima et al.  completed a cost-effectiveness analysis using the REAL study in Canada and found brand risedronate to be cost-effective relative to brand or generic alendronate. This study provides a good starting point for assessing the cost-effectiveness of bisphosphonate therapies utilizing real-world data; however, the applicability of the conclusions may be limited for other countries due to the use of Canadian epidemiology and cost data. Typically, the generalizability of health economic evaluations is cautioned due to factors such as health-care resources and relative prices that vary from region to region that might alter the cost-effectiveness result [29,30].
Given the predicted increases in the burden of osteoporosis in Germany, as well as the reported low awareness of this disease, it is essential to evaluate the health economics of the most effective therapies with the best available evidence. Therefore, utilizing the unique effectiveness data available from the REAL study, as well as German-specific epidemiology and cost data, the objective of this study was to examine the cost-effectiveness of risedronate compared to generic alendronate in the treatment of postmenopausal women with osteoporosis in Germany.
A published Markov state-transition model  was adapted to Germany to compare brand risedronate with generic alendronate in the treatment of postmenopausal women 65 years of age or older with osteoporosis (BMD T-score ≤2.5). Generic alendronate was selected as the comparator because of its lower price compared to branded alendronate in Germany. The perspective for this analysis was the German Social Insurance (GSI) system that encompasses several health-care sectors: treatment and acute hospitalization covered by statutory health insurance (SHI, Krankenversicherung), rehabilitation covered by pension insurance and long-term care (LTC) covered by long-term care insurance (Pflegeversicherung).
The model used in the current study is consistent with an osteoporosis reference model proposed by Zethraeus et al.  and has been used in prior analyses of risedronate in Germany and other countries globally [33–36]. The validity and technical accuracy of the model has been well-documented and these details are provided elsewhere [31–37].
The model is a state-transition cohort model that uses a 1-year cycle length and a fracture incidence-based approach to simulate the occurrence of fractures and the impact of therapy. At the start of the model, a cohort of patients begins in the “well” state, and each year remains in the “well” state or moves to alternative health states according to state- and age-dependent transition probabilities (see Fig. 1). Patients who experience a hip fracture transition to the “1st hip fracture” state and remain in this state for the year. At the end of the year, patients transition to either the “post–1st hip fracture” state or the “dead” state, depending on the transition probabilities. Once in the “post–1st hip fracture” state, patients may remain in that state, experience a 2nd hip fracture, or die due to causes other than hip fracture. Patients surviving a second hip fracture enter the “post–2nd hip fracture” state where they either remain or die due to other (non-fracture related) causes.
Each health state is associated with an economic cost and a health utility weight, which are used to estimate the hip fracture-related costs and quality-adjusted life years (QALYs), respectively. The post–hip fracture states capture the long-term cost and quality-of-life impacts of chronic morbidity associated with hip fractures. Annual cost and QALY data for each year in the model were summed to estimate the total outcomes for the cohort.
The population for this analysis included women 65 years of age or older with osteoporosis (with or without a previous vertebral fracture) and a BMD T-score of ≤−2.5 and reflected the broad range of patients for which bisphosphonates are recommended as a treatment option in Germany . A population analysis was completed for Germany using German-specific data for the age-specific prevalence of osteoporosis and age-specific prevalence of existing vertebral fracture [1,39].
The analysis considered eight cohorts with unique risk profiles, defined according to: 1) age (65–69, 70–74, 75–79, and 80+); and 2) presence or absence of a prevalent vertebral fracture, as both age and vertebral fracture are strong predictors of future fracture . The costs, fractures, and QALYs calculated for each of the eight cohorts were weighted based on the estimated proportion of the population that comprised each cohort in Germany. These proportions were derived using general population data from the Federal Statistical Office (Statistisches Bundesamt)  and combined with the estimated prevalence of osteoporosis in each age group for Germany  (see Table 1). This was then utilized in conjunction with data on the age-specific prevalence of existing vertebral fractures in the female population in Germany . Because data on the prevalence of vertebral fractures in women 80+ years of age were not available, the proportion was assumed to be equivalent to that reported for women 75 to 79 years of age.
Table 1. Distribution of women among the eight cohorts defined by age and previous vertebral fracture for the model based on German-specific prevalence data*
Proportion of osteoporotic population over 65 years (%)
Previous vertebral fracture (%)
Age-specific breakdown of osteoporosis proportions based on data provided from a German sickness fund covering 1.5 million beneficiaries and billing data for outpatient visits that estimated osteoporosis prevalence in Germany . Prevalence of vertebral fracture was based on a multicenter, cross-sectional population-based survey of 1916 women in Germany .
To further adapt the model to capture Germany-specific fracture risk, age-specific incidence of hip fracture for 2005 to 2006 from Federal Statistical Office was utilized  (Table 2). Although these data were specific to women, they were not specific to women with postmenopausal osteoporosis, the target population for this analysis. These general population fracture estimates were, therefore, adjusted to reflect the increased risk of fracture in women with low BMD using a methodology reported by Black et al. 2006 . The relative risk values that were used in the base-case analysis are presented in Table 3 and compared to an independent study by Kanis et al.  that reports similar values. Utilizing these age-specific relative risk values, the published general population fracture incidence rates reported in Germany were adjusted to obtain the estimated fracture incidence rates for the target population of the analysis (i.e., German postmenopausal women with osteoporosis, with or without previous vertebral fracture).
Table 2. Summary of key model input data
The costs, captured in 1995 and 1996, were updated to 2008 Euros using the medical care component of the Consumer Price Index for Germany (Federal Statistical Office (Statistisches Bundesamt) [Germany]).
BMD, bone mineral density.
All-cause annual mortality (per 1000 patients)
Federal Statistical Office (Statistisches Bundesamt) [Germany]
Relative risk of mortality in the year following hip fracture
Relative risk multiplied by general population mortality to obtain mortality associated with hip fracture in German population. These values then adjusted downward 70%, based on Zethraeus et al. , to avoid overestimating the beneficial effects of treatment on mortality (Kanis et al. )
Hip fracture incidence in the German female population (per 1000)
Federal Statistical Office (Statistisches Bundesamt) [Germany]
Relative risk of hip fracture in women aged ≥ 65 and BMD ≤ −2.5
The age-specific probabilities of death in the general population were derived from Federal Statistical Office . To estimate hip fracture mortality, the relative risk of mortality in the year after hip fracture was obtained from Kanis et al.  and multiplied by the German-specific general population mortality to obtain the probability of death associated with hip fracture. There have been a number of publications, however, which have reported that not all deaths associated with hip fracture are attributable to the fracture and thus preventable by averting the fracture [43–45]. In the recently published osteoporosis “reference model,” only 30% of deaths occurring in the year after the hip fracture were assumed to be causally related to the fracture . Therefore, in the analysis, excess mortality in the year after the hip fracture was reduced by 70% to avoid overestimating the impact of fracture on associated mortality. In the base-case, it was assumed that there was only an excess mortality associated with hip fracture for the first year, as the excess mortality is greatest immediately after the fracture [44,45] (Table 2).
The REAL study, which assessed the incidence of hip and nonvertebral fractures more than 12 months after the initiation of risedronate or alendronate therapy, provided the effectiveness data for the model . The study included 33,830 women 65 years of age or older from commercially available datasets of health-care utilization (from the 1 health plan within Ingenix Lab/RX Database and more than 100 health plans of employers within the Marketscan® Commercial Claims and Encounter database) who initiated once-a-week risedronate or alendronate between July 2002 and September 2004. The study found that patients treated with risedronate had an 18% (95% CI: 0.02–0.32) lower incidence of nonvertebral fractures and a 43% (95% CI: 0.13–0.63) lower incidence of hip fractures at 12 months, compared to alendronate.
In addition to a direct comparison of risedronate and alendronate, both the risedronate and alendronate treatment arms were compared with 3060 “no therapy” patients in the REAL study . “No therapy” patients were defined as those who received only one prescription for either risedronate or alendronate and no additional bisphosphonate use, and their reported hip fracture incidence was 0.8% . This incidence was comparable to the incidence rates calculated at 12 months for the placebo group of the risedronate RCT (i.e., 1.1%) , as well as the alendronate RCT (i.e., 0.7%) . The REAL study found a relative risk of 0.493 (P = 0.01) for hip fracture for risedronate compared to no therapy and 0.882 (P = 0.59) for alendronate compared to no therapy, over a 12-month period . In the model, the adjusted age-specific fracture rates in the “no therapy” group were multiplied by the fracture reduction measures to estimate fracture rates in the treated cohorts.
One year of treatment with risedronate and alendronate was assumed, as this was the observation period of the REAL study from which the treatment effectiveness data were derived. It was further assumed that there was no discontinuation over the 1-year treatment period for comparison to other osteoporosis models that frequently use this assumption, although a published discontinuation pattern specific to Germany was assumed in a sensitivity analysis. For simplicity, the model assumed no residual effect; upon discontinuation of treatment at 1 year, the fracture risk returned immediately to that of osteoporotic women with no treatment. It was also assumed that generic alendronate was equivalent in efficacy to branded alendronate.
Health Utility and Costs
Health utility values for the general population were not available for Germany. Thus, for patients in the “well” state (e.g., patients that have not fractured) the calculation of QALYs was based on age-specific utility values published for Sweden  (see Table 2). For patients in the “post hip fracture” state, utility values were derived by applying a multiplier of 0.792  to the age-specific general population utility values. In the following year, a utility multiplier of 0.813 was applied, and in all subsequent years 3+, the utility multiplier was assumed to increase to 0.90 for patients in the “post hip fracture” state [32,48].
Annual German-specific drug costs for risedronate (€507.35) and generic alendronate (€273.39) were obtained from Lauer-Taxe online  (Table 2). Annual drug costs for risedronate and alendronate were calculated using the following formulas (1.39 × 365 = 507.35, 0.7490 × 365 = 273.39). A patient co-payment value of €43.33 was subtracted from the risedronate cost given that the GSI would reimburse this value. There is no copayment for generic alendronate. The costs of hip fracture in the first year after the fracture and in each year subsequent to the fracture were taken from a published German study . The costs, captured in 1995 to 1996, were updated to 2008 Euros using the medical care component of the Consumer Price Index (CPI) for Germany  (see Table 2).
We conducted analyses to estimate the clinical and economic impact of treatment with risedronate compared to generic alendronate in the German population of postmenopausal women 65 years of age and older with osteoporosis. Patients were treated for 1 year and followed in the model for an additional 4 years to capture the ongoing impact of the difference in fractures during the first year on costs, life years and QALYs in subsequent years. Although guidelines for economic evaluations often recommend a lifetime horizon where the disease or therapy has a chronic impact, a shorter time horizon was considered based on the 1-year duration of the REAL study. A cost-effectiveness analysis was performed using the incremental cost per hip fracture avoided and the cost per QALY gained. Costs and outcomes were discounted at a rate of 3%.
The analysis focused on hip fracture alone as hip fractures have the greatest clinical (morbidity and mortality) and economic consequences (costs) of all osteoporotic fractures. Vertebral fractures were not considered as the effectiveness data were not available from REAL; however, it should be noted that clinical data do not suggest differences in efficacy between alendronate and risedronate in vertebral fracture risk [4,9]. Although the REAL study did provide effectiveness data for nonvertebral fractures that showed a benefit of risedronate over alendronate, comprehensive data for nonvertebral fracture costs and incidence rates are not available for Germany and thus not included in the analysis.
Probabilistic Sensitivity Analysis
The analysis incorporated a probabilistic sensitivity analysis (PSA) using the 95% CI for the relative risk of fracture for risedronate (range 0.28 to 0.86) and alendronate (range 0.56 to 1.40). A triangular distribution was used to characterize the relative risk of fracture associated with risedronate and alendronate treatment to allow for relative risk reductions in a negative range. A lognormal distribution could not be used for alendronate, as the relative risk reduction in the model would cross 0. Fracture costs were varied as well, using a normal distribution for year of hip fracture (range €13,745 to €35,414) and each subsequent year (range €3455 to €20,878). Lower bound estimates were based on the lowest observed cost data reported in Germany. Upper bound estimates were derived by applying the same percentage difference between the lower bound and base-case fracture costs. Crystal Ball Pro 7.3 (2007) software (Oracle Corporation, Redwood Shores, CA) was used to simulate a total of 1000 model runs. For each model run, a random number was selected for each distribution, assuming independence of distributions, and a relative risk value from the distributions was selected for each therapy. The model was then run using the selected relative risk value and a cost-effectiveness ratio calculated. The cost-effectiveness results for the 1000 model runs were combined to determine: 1) the proportion of runs in which risedronate was dominant (lower costs and better outcomes); 2) the proportion of runs in which generic alendronate was dominant; 3) the proportion of runs in which risedronate was more effective and more costly; and 4) the proportion of runs in which risedronate was less effective but less costly. For option 3, the cost-effectiveness ratios of risedronate versus alendronate were calculated. Results were then summarized as the proportion of runs among 1 to 3 in which risedronate was dominant or the cost per fracture averted/cost per QALY gained fell below acceptable thresholds of €5000, €10,000, €15,000, €20,000, €25,000, and €30,000.
Other Sensitivity Analyses
One-way deterministic sensitivity analyses were also performed to determine the impact on the base-case results of uncertainty in key model inputs:
1Two years of treatment instead of 1 year was assumed, with effectiveness of risedronate and alendronate equal in year 2 (e.g., 0.493 relative risk vs. no therapy for both risedronate and alendronate).
2Regarding hip fracture mortality, two unique methodological assumptions were made to further assess the variability reported in the literature:
• assuming that 100% of associated hip fracture mortality is caused by the hip fracture; and
• hip fracture mortality risk extending beyond the first year, up to 5 years based on a recent data source .
3For treatment discontinuation, German-specific data were used for discontinuation rate of therapy (i.e., 54% by 12 months, where the majority (33%) discontinued by the fourth month) . A hypothetical analysis incorporating half of this discontinuation rate (i.e., 27%) was also completed.
4Hip fracture-related costs were lowered utilizing German specific data ; however, it should be noted that detailed costs specific to hip fractures were not comprehensive and thus the hip fracture-related costs from these data may be underestimated. These German costs were €13,745 and €3455 for year of hip fracture and each subsequent year respectively.
5Hip fracture incidence rates were lowered using an alternate German data source .
6The price of generic alendronate is predicted to decrease in Germany, thus we reduced the price by 10% and 30%.
7Utility values were assumed to be different for women with and without a vertebral fracture. Specifically, for women with a previous vertebral fracture, the starting general population utility values were adjusted downward using a multiplier of 0.913 .
8Analyses were conducted from the SHI perspective that includes first year treatment and hospitalization costs but excludes LTC costs in first and subsequent years.
9A discount rate of 5% was used on costs and outcomes.
Finally, given the high likelihood of a future alendronate price reduction, as well as the uncertainty associated with a limited 1-year treatment period, we conducted a two-way sensitivity analysis incorporating a 2-year treatment period and maximum predicted alendronate price reduction of 30%.
Over the 1-year treatment period and 4-year follow-up period, treatment with risedronate resulted in fewer fractures and additional QALYs relative to treatment with alendronate (Table 4). Eight additional hip fractures were averted and 4.1 QALYs were gained (discounted values) per 1000 women treated with risedronate versus alendronate. Under the base-case scenario, treatment with risedronate was also cost-saving: the net present value of the associated cost savings from the perspective of the GSI was €278 per woman treated with risedronate versus alendronate. Risedronate, therefore, provided additional fracture protection at a lower cost than alendronate.
Table 4. Results (per 1000 patients) for two German perspectives (GSI, SHI) for women ≥ 65 years of age with BMD ≤ −2.5 (with or without previous vertebral fracture) treated for 1 year, with a time horizon of 5 years
BMD, bone mineral density; GSI, German Social Insurance; QALY; quality-adjusted life years; SHI, Statutory Health Insurance.
In more than 79% of the PSA simulations, risedronate remained the dominant therapy, as it was less costly and more effective compared to generic alendronate. In more than 80% of the simulations, the cost per fracture averted and the cost per QALY gained for risedronate compared to generic alendronate was below €30,000 (Fig. 2).
The base-case analysis assumed the perspective of the GSI. When the perspective of the SHI only was adopted, risedronate remained the most effective therapy in terms of fractures averted and QALYs gained; however, it was no longer the less costly therapy. The analysis resulted in a cost per fracture averted of €11,016 and a cost per QALY gained of €21,486 (Table 4). Similarly, the direction of the result changed, but to a lesser extent, when varying the data source for hip fracture costs (Table 5).
Table 5. Sensitivity analysis results per 1000 osteoporotic patients in Germany. Results are expressed for risedronate relative to alendronate
Total cost savings
Total hip fractures averted
Total QALYs gained
Cost per hip fracture averted
Cost per QALY gained
LTC, long-term care; QALY; quality-adjusted life years.
Treatment duration = 2 years
Mortality post–hip fracture
1. 100% deaths caused by fracture
2. Excess mortality persists up to 5 years
Discontinuation of Therapy:
1. 27% by 12 months (Assumption)
2. 54% by 12 months (German data)
Fracture incidence: Lower rates (German data)
Utility: differs by vertebral fracture presence
Fracture costs: lower values (German data)
SHI perspective (excludes LTC costs)
Potential alendronate price reductions:
1. Decrease by 10%
2. Decrease by 30%
Treatment = 2 years, 30% price decrease
For all remaining one-way sensitivity analyses (Table 5), risedronate remained the dominant therapy compared to generic alendronate. In terms of the cost savings attributable to risedronate, the base-case results were most sensitive to alternative assumptions regarding treatment duration, therapy discontinuation, and cost of generic alendronate.
When the treatment duration was increased to 2 years and the effectiveness of risedronate compared to generic alendronate was equal in year two, the cost savings attributable to risedronate decreased from the base case of €278 to €106 per patient. When utilizing recent German data on discontinuation rates of bisphosphonates (i.e., 54% by 12 months), results still remained cost saving but the associated cost savings with risedronate decreased by approximately 40% to €162 per patient. Due to uncertainty in future drug prices in Germany, a sensitivity analysis was conducted that considered the impact of lowering the cost of alendronate. When the cost associated with generic alendronate was decreased by 10%, the cost savings associated with risedronate decreased by approximately 10% to €251 per patient. When the cost of alendronate was decreased by 30%, cost savings decreased by approximately 30% to €197 per patient.
Alternative assumptions regarding discount rate, post–hip fracture mortality and utility resulted in only minor increases or decreases in the cost savings associated with risedronate. In none of these analyses did the estimated cost savings fluctuate by more than 10%. One interesting finding was when assuming higher hip fracture-related mortality (i.e., either more deaths caused by hip fracture or hip fracture mortality persisting beyond 1 year), there were more QALYs gained with risedronate therapy.
Lastly, Table 5 additionally describes the results of a two-way sensitivity analysis that was conducted varying the alendronate price reduction to 30% and assuming a 2-year treatment duration. Results show that risedronate is no longer cost savings, but remains cost-effective when compared to alendronate.
The recently published results of the REAL study, which quantified the comparative effectiveness of risedronate and alendronate in actual clinical practice, provide an opportunity to explore the cost-effectiveness of risedronate versus generic alendronate using real-world data. We applied these data in an analysis of women typically treated in Germany for osteoporosis, by simulating multiple cohorts of women 65 years of age or older with a BMD ≤−2.5, with or without a previous vertebral fracture. The base-case analysis considered the perspective of the GSI, and results showed that therapy with risedronate versus alendronate resulted in fewer hip fractures and more QALYs, at a reduced cost. The results reflect the larger 12-month fracture reduction for risedronate compared to alendronate observed in the REAL study. Further, when incorporating the uncertainty in the effectiveness of risedronate and alendronate, the PSA analysis showed that risedronate was the dominant therapy in close to 80% of simulations.
The strength of the present analysis is in its use of comparative effectiveness data from the REAL study. Previous economic evaluations involving alendronate and risedronate have mainly relied on combining efficacy data from noncomparative clinical trials that were heterogeneous in terms of patient inclusion/exclusion criteria, baseline demographics, clinical assessment of fracture outcomes and statistical approach. Only one published economic evaluation utilized the REAL data for comparison of alendronate and risedronate in a Canadian analysis . Results were comparable to our German study in that more fractures were averted and more QALYs were gained with risedronate relative to alendronate therapy. Nevertheless, although the Canadian analysis showed a favorable incremental cost-effectiveness ratio (i.e., $3877 per QALY gained) for risedronate compared to generic alendronate, the current German study results are unique in that cost savings were achieved in the base-case and a majority of the sensitivity analyses. This difference shows that factors affecting the results extend beyond the effectiveness data, and are likely attributable to region-specific parameters, such as fracture incidence and treatment costs.
As described, the use of effectiveness data in an economic evaluation is preferable to efficacy data, and ideally these data would be available from a German population, thus the lack of German-specific effectiveness data is a limitation of the study. Nevertheless, given published US census and prevalence data [53,54], it is estimated that the majority of the osteoporotic population 65+ in the United States is Caucasian and thus are racially similar to Germany. There are also a few important factors noted in the REAL study that would increase confidence in generalizability: 1) the population within the REAL study was very large (i.e., >35,000 patients) and drawn from multiple health plans in many US states; 2) the population consists of subjects with a wide mixture of health characteristics; 3) observations of fracture rates in the database are consistent with clinical trial data; and 4) the length of observation of therapy adherence was consistent to previously reported average duration of adherence to bisphosphonates. The population characteristics of the REAL study (i.e., women more than 65 years, new users of bisphosphonates, exclusion of secondary causes of osteoporosis) suggest that the population is comparable to our modeled study; however, with one important caveat: the REAL study most likely included a proportion of women with higher BMD (i.e., osteopenia) and therefore may reflect a lower risk population. There is some evidence that suggests differences in treatment effect with different osteoporosis risk groups ; however, the implications of this are uncertain at this time given lack of complete data on fracture history and bone density levels in the REAL study. Thus, when extending these results to Germany, we must note that our study modeled a specific population at risk of postmenopausal women aged 65 years or older with osteoporosis (with or without vertebral fractures) and the results of our analysis would be most applicable to this high-risk population in Germany.
It is important to point out the methodological limitations of the REAL study that are common to observational studies. In particular, there is the possibility that systematic errors (e.g., selection bias, measurement misclassification) contribute to observed differences. Although selection bias may arise from differences in fracture risk between the two cohorts of patients at the start of therapy, the near unity in fracture incidence between the two cohorts during the first 3 months of therapy suggest that both cohorts had similar risk for fracture at the initiation of therapy. Misclassification of fracture events and therapy use within health-care utilization data are inevitable; however, given previous research on the accuracy of fracture claims, it is likely that there may be less misclassification with hip fracture outcomes. These study limitations were addressed in detail in the original publication and explored extensively through sensitivity analyses .
The analysis is subject to the limitations imposed by the use of any decision modeling technique. We employed conservative assumptions in the base-case analysis, such that the effectiveness difference between risedronate and alendronate would be based on the effectiveness data in the first year. For example: 1) we excluded residual effect of therapy after stopping treatment—efficacy is often assumed to decline linearly over a period equal to the treatment period; and 2) time horizon was limited to 5 years even though benefits of fracture avoidance would continue after that point. In addition, uncertainty in underlying model assumptions and key model parameters were addressed, where feasible, through probabilistic sensitivity and one-/two-way deterministic sensitivity analyses. The magnitude of the cost savings observed in the base-case was sensitive to alternative assumptions regarding treatment duration, therapy discontinuation and cost of generic alendronate. Results were however most sensitive to assumptions that reflect differences in fracture treatment costs (i.e., lower fracture treatment costs, SHI perspective that excludes LTC costs), changing the results from the direction of cost-savings to incremental cost-effectiveness. Based on these sensitivity analyses, results were still cost-effective by international standards with the highest ratio reported to be €21,486 per QALY .
The base-case analysis was limited to 12 months of treatment and short-term follow-up to reflect the REAL study. As long-term follow-up data become available, the impact of long-term treatment with risedronate and alendronate may be investigated. Longer-term data in the second year of treatment may show a smaller difference in efficacy between the two products. Notably, even under the assumption that the products have equal efficacy in year two, risedronate remains cost saving (€106), implying that the value of early fracture protection is €106 if patients are treated for 2 years.
The results of the analysis are consistent with published cost-effectiveness analyses of risedronate in Germany. Brecht et al. (2003)  found risedronate to be cost-effective in an analysis versus the current standard treatment in Germany for women at high risk of osteoporotic fracture because of low BMD (T-score ≤ −2.5) and a history of previous vertebral fracture. From the perspective of the GSI and over a 3-year treatment period and a 10-year time horizon, risedronate dominated the current average treatment in Germany with net cost savings of €340 per treated woman. From the perspective of the SHI, treatment with risedronate resulted in a cost per hip fracture averted of €33,856 and a cost per QALY gained of €35,690. In a similar modeling study , the authors conducted a comparative cost-effectiveness analysis of risedronate, alendronate, raloxifene, and “no therapy” for German patients at high risk of osteoporotic fracture because of low BMD (T-score ≤ −2.5) and a history of at least one previous vertebral fracture. The base-case analysis was conducted on a cohort of 1000 women 70 years of age, more than 3 years of treatment, and a 10-year time horizon. Treatment with risedronate was cost saving compared to alendronate and raloxifene. In the Brecht study , brand alendronate prices were used, unlike the current analysis that used generic alendronate prices. Thus, it is important to note that when incorporating real-world effectiveness data and the lower generic alendronate prices, our results still show the favorable cost-effectiveness of risedronate relative to alendronate. The current analysis is limited in that it does not incorporate raloxifene therapy because the model does not cover the extraskeletal effects of such therapy. Nevertheless, bisphosphonates are increasingly being used as first-line therapies in this population. Future analyses will however need to consider comparisons to once-yearly intravenous bisphosphonates, once more data become available.
When the REAL study effectiveness data are considered in a cost-effectiveness framework, cost savings are indicated, providing strong support for the use of risedronate compared to generic alendronate for the treatment of osteoporotic German women, 65 years of age or older at risk of fracture.
Funding for this study was provided by the Alliance for Better Bone Health (P&G Pharmaceuticals, Inc. and Sanofi-Aventis). We gratefully acknowledge consultation regarding payer perspective and components of hip fracture costs in Germany from Josef Georg Brecht, Scientific Director and President of InForMed GmbH—Outcomes Research and Health Economics.
Source of financial support: Melissa Thompson and Daniel Grima are employees of and shareholders in Cornerstone Research Group Inc. Cornerstone received funding to conduct the cost-effectiveness analysis and develop the article. Margaret Pasquale and Werner Moehrke are employees of P&G Pharmaceuticals Inc. Hans Peter Kruse is an advisory board member of P&G Pharmaceuticals-Germany.